From owner-freebsd-hackers Tue Jul 30 11:25:34 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id LAA24287 for hackers-outgoing; Tue, 30 Jul 1996 11:25:34 -0700 (PDT) Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id LAA24282 for ; Tue, 30 Jul 1996 11:25:30 -0700 (PDT) Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id LAA00292; Tue, 30 Jul 1996 11:21:45 -0700 From: Terry Lambert Message-Id: <199607301821.LAA00292@phaeton.artisoft.com> Subject: Re: Ethernet-like loopback & IPX To: louie@TransSys.COM (Louis A. Mamakos) Date: Tue, 30 Jul 1996 11:21:45 -0700 (MST) Cc: babkin@hq.icb.chel.su, jhay@mikom.csir.co.za, hackers@FreeBSD.org In-Reply-To: <199607230410.AAA24684@whizzo.transsys.com> from "Louis A. Mamakos" at Jul 23, 96 00:10:32 am X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-hackers@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk > > > > > ... > > > > > > > > > > By the way, I have fixed the bug (in my opinion) in IPX. It allowed > > > > > only one link-level (i.e. Ethernet) address for all interfaces. > > > > > Was it a real bug or feature ? > > > > > > > > > > > > > It is a leftover from its XNS days. If you send me the patches for > > > > IPX, I will look it over and commit it. > > > > > > It's not a bug, but a feature. Having a single address for your > > > end-station make quite a few other weirdnesses in the protocol stack > > > and application just go away. > > > > But if you have a single address you need to set it for all > > Ethernet cards or they will not accept the packets with this > > address. The current implementation does not do this and I > > think that it's not a very good idea. > > Sure, it's true that the current drivers do not allow the MAC address > to be set on the various LAN interfaces, but this would argue for > fixing the device drivers, and not the protocol design. > > While you may believe that the single address per IPX host is an > error, it is fundamental to the protocol's design. I wouldn't > characterize changing this as "fixing" it, unless it's the same way a > veteranarian "fixes" something. You are mistaken. Probably, you don't know about "internal net addresses", which were introduced in NetWare 3.x. Each NetWare server from 3.x onward (and thus each router) has the concept of an internal net address. This is basically an internal virtual network interface. The destination is the internal interconnect address, which is not a real hardware address. Conceptually, it looks like this: Server address | o--x---x-------------------x--------------------x--o | | | | | | Ethernet 1 Ethernet 2 Ethernet 3 The destination address specified is "Server address" in all cases. Each ethernet interface has a logical name. This logical name is the same for each IPX machine's attachment to that interface. The practical effect is that the hop limit is reduced by one (from a default 16 to a default 15) by the addition of the routing from the internal to the external net interfaces. Thus you have a unique "server address" and you have a unique net address per card, but each card's net address is not unique. This is pretty obvious if you look at the interface configuration on the IPX stack on a UnixWare box, or on a NetWare 3.x/4.x server. You should contact Jim Freeman of Caldera, since he worked on the IPX stack which was shipped in UnixWare, if you don't believe this. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.